doi 10.1098/rspb.2001.1841

Genetic evidence for extreme polyandry and extraordinary sex-role reversal in a pipeŽ sh Adam G. Jones1*, DeEtte Walker2 and John C. Avise2 1Department of Zoology, 3029 Cordley Hall, Oregon State University, Corvallis, OR 97331, USA 2Department of Genetics,The University of Georgia, Athens, GA 30602, USA Due to the phenomenon of male pregnancy, the ¢sh family Syngnathidae (seahorses and pipe¢shes) has historically been considered an archetypal example of a group in which sexual selection should act more strongly on females than on males. However, more recent work has called into question the idea that all species with male pregnancy are sex-role reversed with respect to the intensity of sexual selection. Furthermore, no studies have formally quanti¢ed the opportunity for sexual selection in any natural breeding assemblage of pipe¢shes or seahorses in order to demonstrate conclusively that sexual selection acts most strongly on females. Here, we use a DNA-based study of parentage in the Gulf pipe¢sh Syngnathus scovelli in order to show that sexual selection indeed acts more strongly on females than on males in this species. Moreover, the Gulf pipe¢sh exhibits classical polyandry with the greatest asymmetry in reproductive roles (as quanti¢ed by variances in mating success) between males and females yet documented in any system. Thus, the intensity of sexual selection on females in pipe¢sh rivals that of any other taxon yet studied. Keywords: mating system; microsatellites; paternity; parentage; sexual selection

typhle (Berglund et al. 1988, 1989; Berglund & Rosenqvist 1. INTRODUCTION 1990, 1993; Vincent et al. 1994). Observations are consist- Male pregnancy is the rule for ¢shes of the family Syng- ent with the idea that females experience greater sexual nathidae (pipe¢shes, seahorses and sea dragons). During selection than males (Vincent et al. 1994), but we copulation, the female transfers eggs to the ventral currently have no method for translating £uctuating surface of the male, where he fertilizes and carries them operational sex ratios into a metric of sexual selection in until hatching (Berglund et al. 1986a,b). Trivers (1972) and a way that allows quantitative comparison of sexual Williams (1975) believed that such dramatic paternal selection among taxa. investment should result in stronger sexual selection on Several other studies have described some aspects of females than on males in these species. However, recent the genetic mating systems of pipe¢sh in nature (Jones & work has called the generality of this prediction into Avise 1997a,b, 2001; Jones et al. 1999; McCoy et al. 2001), question by showing that seahorses are monogamous and but none have obtained a su¤ciently complete description appear not to be sex-role reversed with respect to the of the mating behaviour of both females and males for direction of sexual selection (Vincent et al. 1992; Vincent permitting a quantitative, comparative evaluation of the 1994; Masonjones & Lewis 1996, 2000; Jones et al. 1998; intensity of sexual selection. Thus, a current challenge is Kvarnemo et al. 2000). Moreover, a careful examination to obtain a su¤ciently complete characterization of the of the syngnathid literature shows that the opportunity natural mating system for a pipe¢sh in order to calculate for sexual selection has not been quanti¢ed for any pipe- commonly used measures of the intensity of sexual select- ¢sh in a way that shows with certainty that sexual selec- ion and compare these results to similar studies of other tion acts most strongly on females in natural populations. sex-role-reversed taxa. Studies of sexual selection in pipe¢sh have mainly An excellent candidate species for additional study of relied on laboratory studies of mating preference genetic mating patterns and sexual selection is the Gulf (Berglund et al. 1986a, 1988; Rosenqvist 1990) or genetic pipe¢sh Syngnathus scovelli. Female Gulf pipe¢sh are larger mating patterns (Jones et al. 2000a,b). Although these than males and have apparent secondary sexual laboratory-based studies have been very informative with characters, such as a deeply keeled abdomen and brilliant respect to some important aspects of the sexual selection stripes, which are absent in males (Brown 1972; Jones & process, they do not provide a de¢nitive measure of the Avise 1997a). These aspects of Gulf pipe¢sh natural history intensity of sexual selection in nature. Two studies have indicate that the females of this species experience strong examined the reproductive ecology of natural pipe¢sh sexual selection. Our goal was to use a microsatellite- populations and provide data on their sex ratios and based assessment of parentage in order to test this movement patterns throughout the breeding season hypothesis and quantify the magnitude of sex-role (Vincent et al. 1994, 1995). These observations can only be reversal for the ¢rst time, to our knowledge, in a natural translated into inferences about the nature of sexual select- population of pipe¢sh. ion when coupled with knowledge of female reproductive rates (Vincent et al. 1995). Such data are available and have been integrated with ¢eld studies only for Syngnathus 2. METHODS Gulf pipe¢sh were collected on 28 July 1996 from a site near *Author for correspondence ( [email protected]). Titusville in the intracoastal waterway on the Atlantic coast of

Proc. R. Soc. Lond. B (2001) 268, 2531^2535 2531 © 2001 The Royal Society Received 8 June 2001 Accepted 15 August 2001 2532 A. G. Jones and others Polyandry and extraordinary sex- role reversal in apipe¢sh

of female mating success. The breeding population size was esti- 0.8 males mated using a mark^recapture framework (Jones & Avise 1997b) in which marked females were considered as those whose females genotypes appeared in the brood pouches of the 27 males. The 0.6 proportion of these genotypes appearing in our sample of adult y c

n females yields an estimate of the number of breeding females in e

u the population (using the Lincoln Peterson method) (Pollock q ^

e 0.4 r

f et al. 1990) of 55.0 11.2 (mean s.d.). Given our estimate of the § § breeding population size, we used a Monte Carlo approach to 0.2 obtain a maximum-likelihood estimate of the distribution of female mating success (a detailed description of this approach 0 has been published elsewhere in Jones & Avise (1997b)). Females were constrained to mate with a maximum of four males (the 0 1 2 3 4 maximum number of mates observed for a female in our number of mates sample), and all possible distributions with a mean mating Figure 1. Distributions of the mating success of male and success of unity (a consequence of the 1:1 ratio of females to female Gulf pipe¢sh. pregnant males) were tested. The top 5% of simulated distribu- tions that most often recovered our observed dataset were aver- aged and used as our estimate of the distribution of female Florida. A small patch of sea grass of ca. 23 m2 in area and mating success for this breeding assemblage. We calculated the surrounded on all sides by at least 2 m of bare, sandy substrate variance in female mating success from this distribution. We ran was sampled exhaustively using a dip net. A total of 34 adult additional simulations with di¡erently sized breeding popu- males and 27 adult females were collected, frozen on dry ice lations and all results were consistent with our main conclusions and returned to the laboratory for genetic analysis. The males (e.g. the maximum-likelihood estimates of the variances in and females were measured and dissected. Ornamentation was female mating success were between 1.6 and 2.1 with population scored using an index from 1 (low intensity) to 5 (high sizes ranging from 35 to 75 females). intensity). The sexual maturity of males was con¢rmed by the presence of a well-developed brood pouch and that of females 3. RESULTS AND DISCUSSION was con¢rmed by the presence of mature ova. Each embryo carried alleles consistent with paternity (a) Microsatellite-based parentage analysis by the male in whose pouch it was found and each clutch Twenty-seven of the 34 adult males were pregnant, 21 of exhibited a maximum of two maternal alleles per locus. which carried broods that were su¤ciently developed for micro- Thus, each male apparently received eggs from a single satellite assay. The broods ranged in size from six to 31 (mean of female, the multilocus genotype of whom could be recon- 16.4) embryos and we assayed nearly every embryo (a mean of structed by subtraction of the paternal alleles from the 14 per male) using four previously described microsatellite loci genotypes of the embryos. We successfully matched the (micro11.1, micro22.3, micro25.10 and micro25.22) (Jones & adult females in our ¢eld sample to 67% of the assayed Avise 1997a). Polymerase chain reaction (PCR) was carried out broods using these reconstructed maternal genotypes. as reported elsewhere (Jones & Avise 1997a) and the PCR frag- Such a high rate of success permits an accurate estimate ments were separated and sized using an ABI377 automated of the breeding population size and the distribution of sequencer. female mating success (see ½ 2). Maternal genotypes were reconstructed from the micro- Our molecular results produced three lines of evidence satellite genotypes of the progeny by subtraction of the known indicating that females experience stronger sexual select- paternal alleles. Simulations run using GERUDsim1.0 (Jones ion than males in this population of Gulf pipe¢sh. First, 2001) showed that both our probability of detecting multiple the average number of mates per mating female was esti- maternities within a brood pouch and our probability of recon- mated to be 2.2, while the average number of mates per structing maternal genotypes correctly exceeded 0.99 for this pregnant male was exactly unity (¢gure 1). Thus, their particular microsatellite marker system. We used the recon- genetic mating system can be described as perfect structed maternal genotypes for matching particular adult classical polyandry, a pattern that is rare in the natural females in our collection to their o¡spring, which were carried world and that is expected to produce strong sexual in the pouches of pregnant males. We successfully identi¢ed the selection on females. Here, we use the term polyandry mothers of 14 out of the 21 broods assayed. The microsatellite (Searcy & Yasukawa 1995) as a population-wide descrip- markers were su¤ciently polymorphic that all individuals had tion of a mating system in which successful females unique multilocus genotypes within our sample, with expected frequently produce o¡spring with several males, whereas frequencies (assuming Hardy^Weinberg equilibrium) ranging males sire o¡spring from at most one female within a from 10710 to 1075. Thus, we can be certain that the genetic well-de¢ned breeding interval (in this case the span of a matches indicate the outcomes of actual mating interactions and single male pregnancy). did not occur simply by chance. Second, the standardized variance in female mating success (the variance divided by the square of the mean), (b) Estimation of the distribution of mating success which indicates the opportunity for sexual selection Our characterization of the mating patterns in this isolated (Wade & Arnold 1980), was at least seven times higher patch of sea grass was su¤ciently complete to allow accurate than the standardized variance in male mating success estimation of the breeding population size and the distribution (calculated with non-pregnant males included). This

Proc. R. Soc. Lond. B (2001) Polyandry and extraordinary sex-role reversal in apipe¢sh A. G. Jones and others 2533

Table 1. A comparison of the phenotypic values for females that were identi¢ed as mates of collected males (n 7) ˆ and females that had not mated successfully within our sample (n 20). ˆ

mated females unmated females t-test measurement mean s.e. mean s.e. p-value snout^vent length (mm) 38.10 0.70 33.80 0.90 0.01 body depth (mm) 3.80 0.20 3.20 0.10 0.02 mass (g) 0.40 0.03 0.29 0.02 0.01 ornamentation (index) 5.00 0.00 3.30 1.20 5 0.001

Table 2. Genetic mating system parameters from some our results are the ¢rst empirical demonstration that of the most extremely polyandrous species ever these same types of characters are used as mating cues in described. natural populations of pipe¢sh. Furthermore, this mating (The data for the spotted sandpiper, bronze-winged jacana, preference is so strong that it is evident in our sample, wattled jacana and red-necked phalarope came from Oring et despite a small sample size. Thus, the sexual selection on al. (1991b, 1992), Butchart (2000), Emlen et al. (1998) and females in this population must be very strong indeed. A Reynolds (1987), respectively.) comparison of the phenotypes of pregnant males with non-pregnant males showed no signal of sexual selection. number of variance variance The two groups of males did not di¡er in snout^ vent mates per in female in male length (t-test, d.f. 10 and p 0.89), body depth (t-test, ˆ ˆ mating mating mating d.f. 8 and p 0.81) or mass (with embryos removed) ˆ ˆ species female success success (t-test, d.f. 8 and p 0.06). ˆ ˆ Other species have been described as classically poly- Gulf pipe¢sh 2.27 2.03 0.17 androus, the most extreme and well-known cases of spotted sandpiper 1.70 1.06 0.17 which occur in shorebirds. A comparison of the genetic bronze-winged jacana 1.57 1.01 0.26 wattled jacana 1.60 0.94 0.15 mating parameters in Gulf pipe¢sh with the mating para- red-necked phalarope 1.10 0.25 0.00 meters of the best characterized of shorebird taxa reveals that the Gulf pipe¢sh exhibits the most extreme form of classical polyandry yet described (table 2). We interpret each individual’s reproductive lifespan in Gulf pipe¢sh as result is particularly robust given that the variance in comprising a series of male pregnancies, each of which male mating success is probably in£ated, as some or all can be counted as a relevant period of time during which of the non-pregnant males in our collection possibly had competition for mates occurs. Thus, male pregnancies are recently given birth and were collected while they were analogous to avian breeding seasons and our results can preparing their brood pouches for a subsequent justi¢ably be compared with single-season studies of pregnancy (i.e. they may have been empty as a con- mating patterns in these other taxa. The mean number of sequence of their reproductive cycle rather than as a mates per mating female during a single male pregnancy consequence of competition for mates). The ratio of in Gulf pipe¢sh is 2.27, a value substantially higher than female-to-male variance in mating success in this popu- the single-season range (1.10^1.70) (table 2) observed for lation is comparable with the ratios of male to female shorebird populations for which the mating system has variances observed in some highly polygynous species been characterized, including the spotted sandpiper Actitis (Clutton-Brock 1988; MacKenzie et al. 1995; Searcy & macularia (Oring et al. 1991b), the wattled jacana Jacana Yasukawa 1995), thereby indicating that the strength of jacana (Emlen et al. 1998), the bronze-winged jacana sexual selection on female pipe¢sh may approach the Metopidius indicus (Butchart 2000) and the red-necked intensity of sexual selection acting on some of the classic phalarope Phalaropus lobatus (Reynolds 1987). examples of dramatic sexual selection on highly In addition, the variance in female mating success, polygynous males. which is directly related to the opportunity for sexual Finally, direct comparison of the phenotypes of the selection, is approximately twice as high in female Gulf females from our sample that had mated successfully with pipe¢sh as compared with sandpipers and jacanas those that were not implicated as the mates of sampled (table 2). The variance in mating success is equal to the males showed that the mating females were a non- opportunity for sexual selection for perfectly polyandrous random subset of the females in the population (table 1). species with equal sex ratios (as both males and females The mating females were larger and more ornamented must have a mean mating success of unity and the oppor- than average, a result indicative of strong sexual selection tunity for sexual selection is the variance in mating on one or more of these characters (depending on the success divided by the square of the mean) (Wade & extent to which they covary). These observations are Arnold 1980). Deviations from an equal sex ratio can consistent with laboratory observations of other pipe¢sh a¡ect the translation of variances into the opportunity for species, in which males prefer to mate with larger and sexual selection, but the e¡ect is small enough that our more ornamented females (Berglund et al. 1986b; conclusions remain valid for a wide range of possible sex Rosenqvist 1993). However, to the authors’ knowledge, ratios. In contrast to the situation for females, successfully

Proc. R. Soc. Lond. B (2001) 2534 A. G. Jones and others Polyandry and extraordinary sex-role reversal in apipe¢sh mating males in all of these species mate with approxi- REFERENCES mately one female and the variances in male mating success are low (table 2). The small amount of variance in Berglund, A. & Rosenqvist, G. 1990 Male limitation of female male mating success that does occur is due to either males reproductive success in a pipe¢sh: e¡ects of body-size di¡er- that fail to mate or the few instances of extra-pair fertil- ences. Behav. Ecol. Sociobiol. 27, 129^133. izations that have been genetically documented for some Berglund, A. & Rosenqvist, G. 1993 Selective males and ardent females in pipe¢shes. Behav. Ecol. Sociobiol. 32, 331^336. of these taxa (Oring et al. 1992; Emlen et al. 1998). In Berglund, A., Rosenqvist, G. & Svensson, I. 1986a Reversed sex short, the Gulf pipe¢sh appears to exhibit the most roles and parental energy investment in zygotes of two pipe- extreme form of classical polyandry yet documented, a ¢sh (Syngnathidae) species. Mar. Ecol. Prog. Ser. 29, 209^215. ¢nding that indicates that this species experiences an Berglund, A., Rosenqvist, G. & Svensson, I. 1986b Mate choice, intensity of sexual selection on females far stronger than fecundity and sexual dimorphism in two pipe¢sh species in any other group studied. (Syngnathidae). Behav. Ecol. Sociobiol. 19, 301^307. Although not all of the studies of shorebird mating Berglund, A., Rosenqvist, G. & Svensson, I. 1988 Multiple systems discussed here used genetic markers for character- matings and paternal brood care in the pipe¢sh Syngnathus izing the mating systems, the rates of cuckoldry in these typhle. Oikos 51, 184^188. organisms appear to be su¤ciently low (Oring et al. Berglund, A., Rosenqvist, G. & Svensson, I. 1989 Reproductive success of females limited by males in two pipe¢sh species. 1992; Delehanty et al. 1998; Emlen et al. 1998) that these Am. Nat. 133, 506^516. intensive within-season studies of reproductive ecology Brown, J. 1972 A comparative life history study of four species of probably reveal an accurate picture of the mating system pipe¢shes (family Syngnathidae) in Florida. PhD dissertation, and, therefore, are directly comparable with our study of University of FloridaöGainesville, FL. the Gulf pipe¢sh. However, studies of lifetime reproduc- Butchart, S. H. M. 2000 Population structure and breeding tive success (e.g. Oring et al. 1991a) would not be compar- system of the sex-role reversed, polyandrous bronze-winged able and this area of research still represents a major void jacana Metopidius indicus. Ibis 142, 93^102. in the study of piscine genetic mating systems. We should Clutton-Brock, T. H. 1988 Reproductive success. University of also note that our study represents a single snapshot of Chicago Press. the Gulf pipe¢sh mating system. Temporal or spatial Dawson, C. E. 1985 Indo-Paci¢c pipe¢shes. Lawrence, KS: Allen variation in the mating system may contribute to the Press. Delehanty, D. J., Fleischer, R. C., Colwell, M. A. & Oring, L.W. dynamics of sexual selection in this species in interesting 1998 Sex-role reversal and the absence of extra-pair fertiliz- ways. Nevertheless, the mating system of this Atlantic ation in Wilson’s phalaropes. Anim. Behav. 55, 995^1002. population of Gulf pipe¢sh appears identical to that of a Emlen, S. T., Wrege, P. H. & Webster, M. S. 1998 Cuckoldry as previously studied population from the Gulf of Mexico, a cost of polyandry in the sex-role-reversed wattled jacana, despite a striking di¡erence in overall body size and Jacana jacana. Proc. R. Soc. Lond. B 265, 2359^2364. (DOI clutch size between the two populations (Jones & Avise 10.1098/rspb.1998.0584.) 1997a). Thus, all available data indicate that genetic poly- Jones, A. G. 2001 GERUD1.0: a computer program for the andry is the prevailing mating system in this species. reconstruction of parental genotypes from progeny arrays Our study vindicates the views of Trivers (1972) and using multi-locus DNA data. Mol. Ecol. Notes 1, 33. Williams (1975) that pregnant male ¢sh of the family Jones, A. G. & Avise, J. C. 1997a Microsatellite analysis of maternity and the mating system in the Gulf pipe¢sh Syngnathidae provide special opportunities for under- Syngnathus scovelli, a species with male pregnancy and sex-role standing the nature of sexual selection. Although some reversal. Mol. Ecol. 6, 203^213. species of seahorse appear not to be sex-role reversed with Jones, A. G. & Avise, J. C. 1997b Polygynandry in the dusky respect to the intensity of sexual selection (Vincent et al. pipe¢sh Syngnathus £oridae revealed by microsatellite DNA 1992; Vincent 1994), our results show that at least some markers. Evolution 51, 1611^1622. pipe¢sh species experience very intense sexual selection Jones, A. G. & Avise, J. C. 2001 Mating systems and sexual selec- on females. Thus, the family Syngnathidae provides tion in male-pregnant pipe¢shes and seahorses: insights from opportunities for the study of sexual selection that are microsatellite-based studies of maternity. J. Hered. 92,150^158. perhaps better than early researchers could have Jones, A. G., Kvarnemo, C., Moore, G. I., Simmons, L. W. & imagined. In a radiation of more than 200 species Avise, J. C. 1998 Microsatellite evidence for monogamy and (Dawson 1985), we see examples of extreme polyandry sex-biased recombination in the Western Australian seahorse Hippocampus angustus. Mol. Ecol. 7, 1497^1505. with very intense sexual selection on females in some Jones, A. G., Rosenqvist, G., Berglund, A. & Avise, J. C. 1999 lineages (this study), perfect genetic monogamy with The genetic mating system of a sex-role-reversed pipe¢sh almost no apparent sexual selection in other lineages (Syngnathus typhle): a molecular inquiry. Behav. Ecol. Sociobiol. (Jones et al. 1998; Kvarnemo et al. 2000) and additional 46, 357^365. mating systems that appear to be intermediate in yet Jones, A. G., Rosenqvist, G., Berglund, A., Arnold, S. J. & other groups (Jones & Avise 1997b; Jones et al. 1999). Avise, J. C. 2000a The Bateman gradient and the cause of Thus, the ¢sh family Syngnathidae represents a largely sexual selection in a sex-role-reversed pipe¢sh. Proc. R. Soc. untapped resource for the study of the evolutionary Lond. B 267, 677^680. (DOI 10.1098/rspb.2000.1055.) causes and consequences of di¡erences between the sexes. Jones, A. G., Rosenqvist, G., Berglund, A. & Avise, J. C. 2000b Mate quality in£uences multiple maternity in the sex-role- reversed pipe¢sh Syngnathus typhle. Oikos 90, 321^326. This work was supported by the National Science Foundation Kvarnemo, C., Moore, G. I., Jones, A. G., Nelson, W. S. & (A.G.J.), funds from the University of Georgia (J.C.A. and D.W.) Avise, J. C. 2000 Monogamous pair bonds and mate and the Pew Foundation (J.C.A.). We thank Ginger E. Carney switching in the Western Australian seahorse Hippocampus for help with the collection of ¢eld samples. subelongatus. J. Evol. Biol. 13, 882^888.

Proc. R. Soc. Lond. B (2001) Polyandry and extraordinary sex-role reversal in apipe¢sh A. G. Jones and others 2535

McCoy, E., Jones, A. G. & Avise, J. C. 2001 The genetic mating Reynolds, J. D. 1987 Mating system and nesting biology of the system and tests for cuckoldry in a pipe¢sh species in which red-necked phalarope Phalaropus lobatus: what constrains poly- males fertilize eggs and brood o¡spring externally. Mol. Ecol. andry? Ibis 129, 225^242. 10, 1793^1800. Rosenqvist, G. 1990 Male mate choice and female^female MacKenzie, A., Reynolds, J. D., Brown, V. J. & Sutherland, competition for mates in the pipe¢sh Nerophis ophidion. Anim. W. J. 1995 Variation in male mating success on leks. Am. Behav. 39, 1110^1115. Nat. 145, 633^652. Rosenqvist, G. 1993 Sex role reversal in a pipe¢sh. Mar. Behav. Masonjones, H. D. & Lewis, S. M. 1996 Courtship behavior Physiol. 23, 219^230. in the dwarf seahorse, Hippocampus zosterae. Copeia 1996, 634^ Searcy, W. A. & Yasukawa, K. 1995 Polygyny and sexual selection in 640. red-winged blackbirds. Princeton University Press. Masonjones, H. D. & Lewis, S. M. 2000 Di¡erences in Trivers, R. L. 1972 Parental investment and sexual selection. potential reproductive rates of male and female In Sexual selection and the descent of man, 1871^1971 (ed. seahorses related to courtship roles. Anim. Behav. 59, B. Campbell), pp. 136^179. Chicago, IL: Aldine. 11^20. Vincent, A. C. J. 1994 Seahorses exhibit conventional sex roles Oring, L. W., Colwell, M. A. & Reed, J. M. 1991a Lifetime in mating competition, despite male pregnancy. Behaviour 128, reproductive success in the spotted sandpiper (Actitis 135^151. macularia): sex di¡erences and variance components. Behav. Vincent, A., AhnesjÎ, I., Berglund, A. & Rosenqvist, G. 1992 Ecol. Sociobiol. 28, 425^432. Pipe¢shes and seahorses: are they all sex role reversed? Trends Oring, L. W., Reed, J. M., Colwell, M. A., Lank, D. B. & Ecol. Evol. 7, 237^241. Maxson, S. J. 1991b Factors regulating annual Vincent, A., AhnesjÎ, I. & Berglund, A. 1994 Operational sex mating success and reproductive success in spotted ratios and behavioural sex di¡erences in a pipe¢sh popula- sandpipers (Actitis macularia). Behav. Ecol. Sociobiol. 28, tion. Behav. Ecol. Sociobiol. 34, 435^442. 433^442. Vincent, A. C. J., Berglund, A. & AhnesjÎ, I. 1995 Oring, L. W., Fleischer, R. C., Reed, J. M. & Marsden, Reproductive ecology of ¢ve pipe¢sh species in one eelgrass K. E. 1992 Cuckoldry through stored sperm in the se- meadow. Environ. Biol. Fish. 44, 347^361. quentially polyandrous spotted sandpiper. Nature 359, Wade, M. J. & Arnold, S. J. 1980 The intensity of sexual select- 631^633. ion in relation to male sexual behavior, female choice and Pollock, K. H., Nichols, J. D., Brownie, C. & Hines, J. E. 1990 sperm precedence. Anim. Behav. 28, 446^461. Statistical inference for capture^recapture experiments. Wildl. Williams, G. C. 1975 Sex and evolution. Princeton University Monogr. 107, 1^97. Press.

Proc. R. Soc. Lond. B (2001)